With the development of automation technology, traditional woolen structure pattern generation methods are often restricted by rules and styles, making it difficult to achieve the requirements of innovation and detail restoration. Therefore, this study proposes an automatic generation algorithm based on a combination of graph neural networks (GNNs) and generative adversarial networks (GANs), aiming to generate high-quality and creative woolen patterns through a deep learning framework. Experimental results show that the proposed GNN-GAN performs well in terms of quality, detail restoration and innovation of generated patterns. GNN-GAN outperforms other algorithms in terms of FID (18.3), SSIM (0.88) and PDI (0.72), and the generated patterns are more delicate and realistic. In addition, GNN-GAN also achieves a good balance in controlling overfitting and reducing computing resource consumption. The automatic generation algorithm for woolen patterns based on GNN and GAN can not only improve the quality and diversity of pattern generation, but also provide more creative design solutions in practical applications, providing technical support for the intelligent design of the wool textile industry.